AUTO-ADMIN: Non-MP3, WAV, OGG, SoundCloud, YouTube, Vimeo, Twitter and Facebook links in this post have been protected automatically. Once the member reaches 5 posts the links will function as normal.
Hi KVR,I’m releasing Audio Optimizer v1.0 — a fully offline Android app that uses a 25 KB neural model (Tiny AutoFUS) to drive a real-time Biquad parametric EQ, all on-device, no cloud, no internet.
It processes live audio via the microphone and adapts EQ parameters in real time based on spectral content, noise level, and perceived clarity — making it ideal for real-world listening enhancement, industrial sound analysis, or educational DSP demos.
Key Features:Model: tiny_autofus.ptl (25 KB, TorchScript Lite)
DSP: Native C++ (NDK) — Biquad IIR filters, FFT, noise gate
Latency: <15 ms (measured on Snapdragon 665)
Input: Microphone (AudioCaptureService)
Output: Real-time EQ adjustments (local playback enhancement)
Extras: WAV export for calibration & offline analysis
License: MIT
Stack: Kotlin + JNI + CMake
Works with youtube, spotify and others.
How to Test:Download:
audio_optimizer_v1.0.apk
Install (enable Install unknown apps)
Grant mic permission
Launch → speak or play audio → EQ adapts live
If the link appears inactive (KVR auto-protection), copy-paste it into your browser — it’s a direct GitHub release asset.
Use Cases:Personalized audio enhancement
Machine sound diagnostics (pumps, motors, drones)
Edge AI + DSP education
Foundation for future VST/AU mobile plugins
Part of the AZURO AI Platform — focused on autonomous scientific discovery and interpretable edge intelligence.
Feedback from the KVR community on DSP design, latency, or model integration is highly welcome.
— KR / Crete Lab/ https://github.com/Kretski/audio-optimi ... oOptimizer (https://github.com/Kretski/audio-optimizer-android/releases/tag/AudioOptimizer)
p.s. Note: If the GitHub link appears non-clickable, it’s due to KVR’s new-user protection. Simply copy-paste it into your browser — it’s a public MIT-licensed project.